Abstract: Although the applications of low-altitude small unmanned aerial vehicles (UAVs) are expanding, certain operational situations threaten safety. Infrared technology is widely used in anti-UAV systems because of its strong adaptability to complex environments. In this study, the problems of incomplete small-target feature usage and indistinct features in complex backgrounds in the infrared image detection task of UAVs were addressed, and the ISTD-DINO infrared UAV target detection model was proposed. The IFE feature enhancement method was proposed, and spatial and channel attention and multi-branch dilated convolution were adopted to fully mine the spatial information of infrared UAV target features; by fusing information at different levels, long-distance dependencies between multi-scale features were captured; the EIoU bounding box loss function was introduced to improve the detection accuracy of the model during training. Experimental results showed that the AP_0.5 of the ISTD-DINO model reached 97.5%, indicating that ISTD-DINO has significant advantages in the infrared UAV target detection task. ISTD-DINO performed well in identifying small infrared targets across various complex scenarios, providing a technical solution with significant reference value for maintaining airspace order and the healthy development of the low-altitude economy.